Inspiration
Websites like amazon, flipkart, alibaba, ebay receive thousands of new entrants in form of books. To see each of these books manually and label them is a tedious and strenuous task. Thanks to techniques like Computer Vision and AI manual work can now be completely eradicated and we can achieve this with a modicum of effort.
What it does
Here in this project we have taken a dataset of 5 different genres of book from Amazon. The genres are Arts&Photography, Cookbooks, History, Law and Science. The dataset is in the following structure
The folder Image has 3 folders train, test and validation and each of these folders contain 5 folders each corresponding to a category. There are 800 images of each category for training, 180 for validation and 20 for testing.
Given an unseen new input image (cover of the book). The code can judge what genre it belongs to.
How we built it
We used Keras with a tensorflow backend to build the system. The system consists of a convolutional neural net. The current system lies completely on a jupyter notebook but can be easily made into a simple GUI or an App which captures an image and sends it to this code running at the backend and then print the final result.
Challenges we ran into
Keras as a technology was completely new to us. We took this challenge so that we could learn the technology and experiment with the current fad :- "Machine Learning"
Accomplishments that we're proud of
By the end of the project we were able to understand the in's and out's of keras. (pun intended :-p)
What we learned
Team Work, Patience, Dedication and passion can help you achieve things you couldn't even imagine.
What's next for Judging a book by its cover
Scaling it to an android application so that general public could take the picture of the cover of the book and get a myriad of information about it.
Built With
- jupyter
- keras
- python
- tensorflow
Log in or sign up for Devpost to join the conversation.